Home

Lab 3 - Logistic Regression

Student Name: Parthasarathy Rajendiran
Student ID: 8883828

---

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

import warnings
warnings.filterwarnings('ignore')

Data Loading and Exploration

df = pd.read_csv('../../practical_labs/datasets/Lab_3/creditcard.csv')
display(df)

	Time	V1	V2	V3	V4	V5	V6	V7	V8	V9	...	V21	V22	V23	V24	V25	V26	V27	V28	Amount	Class
0	0.0	-1.359807	-0.072781	2.536347	1.378155	-0.338321	0.462388	0.239599	0.098698	0.363787	...	-0.018307	0.277838	-0.110474	0.066928	0.128539	-0.189115	0.133558	-0.021053	149.62	0
1	0.0	1.191857	0.266151	0.166480	0.448154	0.060018	-0.082361	-0.078803	0.085102	-0.255425	...	-0.225775	-0.638672	0.101288	-0.339846	0.167170	0.125895	-0.008983	0.014724	2.69	0
2	1.0	-1.358354	-1.340163	1.773209	0.379780	-0.503198	1.800499	0.791461	0.247676	-1.514654	...	0.247998	0.771679	0.909412	-0.689281	-0.327642	-0.139097	-0.055353	-0.059752	378.66	0
3	1.0	-0.966272	-0.185226	1.792993	-0.863291	-0.010309	1.247203	0.237609	0.377436	-1.387024	...	-0.108300	0.005274	-0.190321	-1.175575	0.647376	-0.221929	0.062723	0.061458	123.50	0
4	2.0	-1.158233	0.877737	1.548718	0.403034	-0.407193	0.095921	0.592941	-0.270533	0.817739	...	-0.009431	0.798278	-0.137458	0.141267	-0.206010	0.502292	0.219422	0.215153	69.99	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
284802	172786.0	-11.881118	10.071785	-9.834783	-2.066656	-5.364473	-2.606837	-4.918215	7.305334	1.914428	...	0.213454	0.111864	1.014480	-0.509348	1.436807	0.250034	0.943651	0.823731	0.77	0
284803	172787.0	-0.732789	-0.055080	2.035030	-0.738589	0.868229	1.058415	0.024330	0.294869	0.584800	...	0.214205	0.924384	0.012463	-1.016226	-0.606624	-0.395255	0.068472	-0.053527	24.79	0
284804	172788.0	1.919565	-0.301254	-3.249640	-0.557828	2.630515	3.031260	-0.296827	0.708417	0.432454	...	0.232045	0.578229	-0.037501	0.640134	0.265745	-0.087371	0.004455	-0.026561	67.88	0
284805	172788.0	-0.240440	0.530483	0.702510	0.689799	-0.377961	0.623708	-0.686180	0.679145	0.392087	...	0.265245	0.800049	-0.163298	0.123205	-0.569159	0.546668	0.108821	0.104533	10.00	0
284806	172792.0	-0.533413	-0.189733	0.703337	-0.506271	-0.012546	-0.649617	1.577006	-0.414650	0.486180	...	0.261057	0.643078	0.376777	0.008797	-0.473649	-0.818267	-0.002415	0.013649	217.00	0

284807 rows × 31 columns

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 284807 entries, 0 to 284806
Data columns (total 31 columns):
 #   Column  Non-Null Count   Dtype  
---  ------  --------------   -----  
 0   Time    284807 non-null  float64
 1   V1      284807 non-null  float64
 2   V2      284807 non-null  float64
 3   V3      284807 non-null  float64
 4   V4      284807 non-null  float64
 5   V5      284807 non-null  float64
 6   V6      284807 non-null  float64
 7   V7      284807 non-null  float64
 8   V8      284807 non-null  float64
 9   V9      284807 non-null  float64
 10  V10     284807 non-null  float64
 11  V11     284807 non-null  float64
 12  V12     284807 non-null  float64
 13  V13     284807 non-null  float64
 14  V14     284807 non-null  float64
 15  V15     284807 non-null  float64
 16  V16     284807 non-null  float64
 17  V17     284807 non-null  float64
 18  V18     284807 non-null  float64
 19  V19     284807 non-null  float64
 20  V20     284807 non-null  float64
 21  V21     284807 non-null  float64
 22  V22     284807 non-null  float64
 23  V23     284807 non-null  float64
 24  V24     284807 non-null  float64
 25  V25     284807 non-null  float64
 26  V26     284807 non-null  float64
 27  V27     284807 non-null  float64
 28  V28     284807 non-null  float64
 29  Amount  284807 non-null  float64
 30  Class   284807 non-null  int64  
dtypes: float64(30), int64(1)
memory usage: 67.4 MB

df.describe(include='all').T

	count	mean	std	min	25%	50%	75%	max
Time	284807.0	9.481386e+04	47488.145955	0.000000	54201.500000	84692.000000	139320.500000	172792.000000
V1	284807.0	1.168375e-15	1.958696	-56.407510	-0.920373	0.018109	1.315642	2.454930
V2	284807.0	3.416908e-16	1.651309	-72.715728	-0.598550	0.065486	0.803724	22.057729
V3	284807.0	-1.379537e-15	1.516255	-48.325589	-0.890365	0.179846	1.027196	9.382558
V4	284807.0	2.074095e-15	1.415869	-5.683171	-0.848640	-0.019847	0.743341	16.875344
V5	284807.0	9.604066e-16	1.380247	-113.743307	-0.691597	-0.054336	0.611926	34.801666
V6	284807.0	1.487313e-15	1.332271	-26.160506	-0.768296	-0.274187	0.398565	73.301626
V7	284807.0	-5.556467e-16	1.237094	-43.557242	-0.554076	0.040103	0.570436	120.589494
V8	284807.0	1.213481e-16	1.194353	-73.216718	-0.208630	0.022358	0.327346	20.007208
V9	284807.0	-2.406331e-15	1.098632	-13.434066	-0.643098	-0.051429	0.597139	15.594995
V10	284807.0	2.239053e-15	1.088850	-24.588262	-0.535426	-0.092917	0.453923	23.745136
V11	284807.0	1.673327e-15	1.020713	-4.797473	-0.762494	-0.032757	0.739593	12.018913
V12	284807.0	-1.247012e-15	0.999201	-18.683715	-0.405571	0.140033	0.618238	7.848392
V13	284807.0	8.190001e-16	0.995274	-5.791881	-0.648539	-0.013568	0.662505	7.126883
V14	284807.0	1.207294e-15	0.958596	-19.214325	-0.425574	0.050601	0.493150	10.526766
V15	284807.0	4.887456e-15	0.915316	-4.498945	-0.582884	0.048072	0.648821	8.877742
V16	284807.0	1.437716e-15	0.876253	-14.129855	-0.468037	0.066413	0.523296	17.315112
V17	284807.0	-3.772171e-16	0.849337	-25.162799	-0.483748	-0.065676	0.399675	9.253526
V18	284807.0	9.564149e-16	0.838176	-9.498746	-0.498850	-0.003636	0.500807	5.041069
V19	284807.0	1.039917e-15	0.814041	-7.213527	-0.456299	0.003735	0.458949	5.591971
V20	284807.0	6.406204e-16	0.770925	-54.497720	-0.211721	-0.062481	0.133041	39.420904
V21	284807.0	1.654067e-16	0.734524	-34.830382	-0.228395	-0.029450	0.186377	27.202839
V22	284807.0	-3.568593e-16	0.725702	-10.933144	-0.542350	0.006782	0.528554	10.503090
V23	284807.0	2.578648e-16	0.624460	-44.807735	-0.161846	-0.011193	0.147642	22.528412
V24	284807.0	4.473266e-15	0.605647	-2.836627	-0.354586	0.040976	0.439527	4.584549
V25	284807.0	5.340915e-16	0.521278	-10.295397	-0.317145	0.016594	0.350716	7.519589
V26	284807.0	1.683437e-15	0.482227	-2.604551	-0.326984	-0.052139	0.240952	3.517346
V27	284807.0	-3.660091e-16	0.403632	-22.565679	-0.070840	0.001342	0.091045	31.612198
V28	284807.0	-1.227390e-16	0.330083	-15.430084	-0.052960	0.011244	0.078280	33.847808
Amount	284807.0	8.834962e+01	250.120109	0.000000	5.600000	22.000000	77.165000	25691.160000
Class	284807.0	1.727486e-03	0.041527	0.000000	0.000000	0.000000	0.000000	1.000000

df['Class'].value_counts()

Class
0    284315
1       492
Name: count, dtype: int64

Data Preprocessing

Outlier Handling

numeric_columns = df.select_dtypes(include='number').columns.to_list()
#Excluding target 
numeric_columns.remove('Class')

Visual inspection of the outliers

fig, axes = plt.subplots(10, 3, figsize=(9, 12))
for idx, col_name in enumerate(numeric_columns):
    i, j = divmod(idx, 3)
    # print(i, j, col_name)
    sns.boxplot(df, x=col_name, ax=axes[i, j])
plt.suptitle('Finding outliers of the features')
fig.tight_layout()
plt.show()

Outlier detection

def is_outlier_whiskers(feature):
    # Finding IQR
    Q1 = feature.quantile(0.25)
    Q3 = feature.quantile(0.75)
    IQR = Q3 - Q1
    lower_fence = Q1 - 1.5*IQR
    upper_fence = Q3 + 1.5*IQR
    
    # Flagging outliers which are apart 1.5 * IQR
    whiskers_outliers = (feature < lower_fence) | (feature > upper_fence)

    return whiskers_outliers

def is_outlier_zscore(feature, threshold=1.5):
    feature_z_score = (feature - feature.mean()) / feature.std()
    z_score_outliers = feature_z_score.abs() > threshold
    return z_score_outliers

Analyzing the z-score to decide the threshold

records = []
for col in numeric_columns:
    feature_z_score = (df[col] - df[col].mean()) / df[col].std()
    records.append([min(df[col]), df[col].mean(), df[col].std(), max(df[col]), min(feature_z_score), feature_z_score.mean(), feature_z_score.std(), max(feature_z_score)])

display(pd.DataFrame(
    records, columns=['Min value', 'Mean value', 'Std.', 'Max value', 'Min z-scr', 'Mean z-scr', 'Std. z-scr', 'Max z-scr'], index=numeric_columns
).round(2))

	Min value	Mean value	Std.	Max value	Min z-scr	Mean z-scr	Std. z-scr	Max z-scr
Time	0.00	94813.86	47488.15	172792.00	-2.00	-0.0	1.0	1.64
V1	-56.41	0.00	1.96	2.45	-28.80	0.0	1.0	1.25
V2	-72.72	0.00	1.65	22.06	-44.04	-0.0	1.0	13.36
V3	-48.33	-0.00	1.52	9.38	-31.87	-0.0	1.0	6.19
V4	-5.68	0.00	1.42	16.88	-4.01	0.0	1.0	11.92
V5	-113.74	0.00	1.38	34.80	-82.41	0.0	1.0	25.21
V6	-26.16	0.00	1.33	73.30	-19.64	0.0	1.0	55.02
V7	-43.56	-0.00	1.24	120.59	-35.21	0.0	1.0	97.48
V8	-73.22	0.00	1.19	20.01	-61.30	-0.0	1.0	16.75
V9	-13.43	-0.00	1.10	15.59	-12.23	0.0	1.0	14.19
V10	-24.59	0.00	1.09	23.75	-22.58	0.0	1.0	21.81
V11	-4.80	0.00	1.02	12.02	-4.70	0.0	1.0	11.78
V12	-18.68	-0.00	1.00	7.85	-18.70	-0.0	1.0	7.85
V13	-5.79	0.00	1.00	7.13	-5.82	0.0	1.0	7.16
V14	-19.21	0.00	0.96	10.53	-20.04	0.0	1.0	10.98
V15	-4.50	0.00	0.92	8.88	-4.92	0.0	1.0	9.70
V16	-14.13	0.00	0.88	17.32	-16.13	0.0	1.0	19.76
V17	-25.16	-0.00	0.85	9.25	-29.63	-0.0	1.0	10.89
V18	-9.50	0.00	0.84	5.04	-11.33	0.0	1.0	6.01
V19	-7.21	0.00	0.81	5.59	-8.86	0.0	1.0	6.87
V20	-54.50	0.00	0.77	39.42	-70.69	-0.0	1.0	51.13
V21	-34.83	0.00	0.73	27.20	-47.42	-0.0	1.0	37.03
V22	-10.93	-0.00	0.73	10.50	-15.07	0.0	1.0	14.47
V23	-44.81	0.00	0.62	22.53	-71.75	0.0	1.0	36.08
V24	-2.84	0.00	0.61	4.58	-4.68	0.0	1.0	7.57
V25	-10.30	0.00	0.52	7.52	-19.75	-0.0	1.0	14.43
V26	-2.60	0.00	0.48	3.52	-5.40	0.0	1.0	7.29
V27	-22.57	-0.00	0.40	31.61	-55.91	0.0	1.0	78.32
V28	-15.43	-0.00	0.33	33.85	-46.75	-0.0	1.0	102.54
Amount	0.00	88.35	250.12	25691.16	-0.35	0.0	1.0	102.36

The z-score of the columns ranging from -82.41 to 102.54.
Usually the threshold would be considered from 2 to 3. Due to the higher range here, we can consider the threshold as 9.

amt_outliers = is_outlier_zscore(df['Amount'], 9)
print(df[amt_outliers]['Class'].value_counts())
df[amt_outliers & (df['Class'] == 1)]

Class
0    505
Name: count, dtype: int64

	Time	V1	V2	V3	V4	V5	V6	V7	V8	V9	...	V21	V22	V23	V24	V25	V26	V27	V28	Amount	Class

0 rows × 31 columns

Outlier Removal

df_raw = df.copy()
for col in numeric_columns:
    outliers = is_outlier_zscore(df[col], 9)
    df = df[~outliers]
print(f'{df_raw.shape[0] - df.shape[0]} outliers have been removed')

4158 outliers have been removed

df['Class'].value_counts()

Class
0    280441
1       208
Name: count, dtype: int64

Normalization

As the Time and Amount columns are in different units than other columns, it is better to normalize the columns. Since the relationship among the variables are important and the outliers are removed, we can use MinMaxScaler to normalize the data.

from sklearn.preprocessing import MinMaxScaler

scaler = MinMaxScaler()
df[numeric_columns] = scaler.fit_transform(df[numeric_columns])
display(df)

	Time	V1	V2	V3	V4	V5	V6	V7	V8	V9	...	V21	V22	V23	V24	V25	V26	V27	V28	Amount	Class
0	0.000000	0.789229	0.535514	0.904386	0.542975	0.483766	0.517793	0.494567	0.506023	0.367281	...	0.498241	0.579649	0.487214	0.423290	0.588957	0.310131	0.523496	0.496004	0.087571	0
1	0.000000	0.930213	0.549180	0.767179	0.471463	0.501649	0.486504	0.476615	0.505309	0.315180	...	0.476850	0.442836	0.512243	0.363989	0.594920	0.368762	0.500174	0.505007	0.001574	0
2	0.000006	0.789309	0.484412	0.860203	0.466205	0.476365	0.594652	0.525683	0.513846	0.209229	...	0.525698	0.653367	0.607760	0.313047	0.518546	0.319441	0.492587	0.486266	0.221625	0
3	0.000006	0.810973	0.530980	0.861348	0.370621	0.498492	0.562872	0.494455	0.520660	0.219968	...	0.488962	0.538962	0.477776	0.242154	0.669038	0.304023	0.511906	0.516767	0.072283	0
4	0.000012	0.800366	0.573839	0.847206	0.467994	0.480675	0.496744	0.514490	0.486634	0.405476	...	0.499156	0.657338	0.484024	0.434127	0.537320	0.438820	0.537545	0.555442	0.040964	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
284802	0.999965	0.207909	0.944547	0.188140	0.278089	0.258131	0.341502	0.203749	0.884452	0.497751	...	0.522136	0.554873	0.620179	0.339279	0.790885	0.391868	0.656040	0.708581	0.000451	0
284803	0.999971	0.823873	0.536227	0.875361	0.380209	0.537931	0.552028	0.482430	0.516324	0.385877	...	0.522214	0.676162	0.501745	0.265384	0.475486	0.271763	0.512847	0.487833	0.014509	0
284804	0.999977	0.970420	0.526302	0.569397	0.394109	0.617044	0.665345	0.464322	0.538040	0.373058	...	0.524053	0.624490	0.495839	0.506854	0.610134	0.329068	0.502373	0.494618	0.039729	0
284805	0.999977	0.851076	0.559838	0.798213	0.490044	0.481987	0.527059	0.442368	0.536503	0.369662	...	0.527476	0.657602	0.480970	0.431494	0.481269	0.447079	0.519449	0.527606	0.005853	0
284806	1.000000	0.834889	0.530798	0.798261	0.398073	0.498391	0.453922	0.569976	0.479066	0.377579	...	0.527045	0.634170	0.544805	0.414816	0.496011	0.193029	0.501249	0.504736	0.127008	0

280649 rows × 31 columns

Descriptive Analytics

Distribution

df['Class'].value_counts()

Class
0    280441
1       208
Name: count, dtype: int64

plt.pie(df['Class'].value_counts(), autopct='%.3f%%', explode=[0, 0.45], labels=['Non-Fraud', 'Fraud'])
plt.title('Distribution of Fraud vs Non-Fraud transactions')
plt.show()

The dataset has very neglible amount (0.074%) of positive class (The fraudulent transactions). The data is highly imbalanced.

Correlation

df_corr = df.corr()
df_corr['Class'].round(2)

Time     -0.00
V1       -0.01
V2        0.02
V3       -0.04
V4        0.05
V5       -0.01
V6       -0.02
V7       -0.03
V8       -0.00
V9       -0.02
V10      -0.06
V11       0.05
V12      -0.08
V13      -0.01
V14      -0.12
V15      -0.00
V16      -0.04
V17      -0.06
V18      -0.01
V19       0.01
V20       0.01
V21       0.02
V22       0.00
V23       0.00
V24      -0.00
V25      -0.00
V26       0.00
V27       0.02
V28       0.01
Amount    0.01
Class     1.00
Name: Class, dtype: float64

Top 10 Correlated Features

top_10_features = df_corr[df_corr['Class'] < 1]['Class'].abs().sort_values(ascending=False)[:10].index
df_corr.loc[top_10_features, 'Class']

V14   -0.118431
V12   -0.075458
V10   -0.064971
V17   -0.059191
V4     0.049861
V11    0.049154
V3    -0.044648
V16   -0.042453
V7    -0.027155
V9    -0.023964
Name: Class, dtype: float64

The V14 feature has more correlation with the Class and it is a Negative Correlation with coefficient -0.118431

df_corr_10 = df_corr.loc[['Class', *top_10_features], ['Class', *top_10_features]]
sns.set_theme(style='white')
mask = np.triu(np.ones_like(df_corr_10, dtype=bool))
cmap = sns.diverging_palette(230, 20, as_cmap=True)

f, ax = plt.subplots(figsize=(15, 6))
sns.heatmap(df_corr_10, mask=mask, cmap=cmap, annot=True, fmt=".2f", ax=ax)
plt.title('Diagonal Correlation Matrix')
plt.show()

# sns.pairplot(df[['Class', *top_10_features]], hue='Class', vars=top_10_features, diag_kind='kde')
# plt.show()

As the pairplot is taking much time to generate, I have plotted it with the test data after train test split.

ML Model Training and Testing

from sklearn.model_selection import train_test_split, cross_validate
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from sklearn.metrics import roc_curve, roc_auc_score
from sklearn.utils.class_weight import compute_class_weight
#from imblearn.over_sampling import SMOTE
from sklearn.pipeline import make_pipeline

Train Test Split

X_train, X_test, y_train, y_test = train_test_split(
    df.drop('Class', axis=1), df['Class'], 
    test_size=0.2, random_state=8883828,
    stratify=df['Class']
)
print(f'The data is splitted into train set with {X_train.shape[0]} records and test set with  {X_test.shape[0]}')

The data is splitted into train set with 224519 records and test set with  56130

Distribution of top 10 features

sns.pairplot(df.loc[X_test.index, ['Class', *top_10_features]], hue='Class', vars=top_10_features, diag_kind='kde')
plt.title('Distribution of top 10 correlated features against the fraud/non-fraud transactions')
plt.show()

Baseline Model

Building Model

baseline_model = LogisticRegression(random_state=8883828)
baseline_model.fit(X_train, y_train)

LogisticRegression(random_state=8883828)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

Learned Coefficients

weights = pd.DataFrame(baseline_model.coef_[0], index=baseline_model.feature_names_in_, columns=['Coefficient'])
display(weights.sort_values(by='Coefficient', key=lambda x:abs(x), ascending=False))

	Coefficient
V14	-8.342693
V4	6.067081
V12	-4.607731
V10	-3.851206
V11	3.610706
V3	-2.863708
V17	-2.319708
V16	-2.301846
V13	-1.786188
V9	-1.758520
V8	-1.505890
V7	-1.460467
V1	-1.430824
V6	-1.344582
Amount	1.332377
V18	-0.637324
Time	-0.567976
V26	-0.550839
V19	-0.535590
V24	0.400604
V28	0.383125
V15	-0.365932
V22	0.355773
V5	0.246263
V2	0.181267
V27	0.171601
V23	0.155047
V25	0.100328
V21	0.090993
V20	0.066082

The model gave maximum weight to V14 and its weight is -8.342693. Next to it, V4 has 6.067081 as its weight.

scoring_metrics = ['accuracy', 'precision', 'recall', 'f1']

cv_results = cross_validate(baseline_model, X_test, y_test, scoring=scoring_metrics, cv=5)
cv_df = pd.DataFrame.from_dict(cv_results)
display(cv_df)
print(cv_df.mean())

	fit_time	score_time	test_accuracy	test_precision	test_recall	test_f1
0	0.267652	0.032116	0.999287	0.0	0.000000	0.000000
1	0.448569	0.036903	0.999555	1.0	0.375000	0.545455
2	0.231677	0.033161	0.999376	1.0	0.125000	0.222222
3	0.384200	0.015100	0.999376	1.0	0.222222	0.363636
4	0.290050	0.029057	0.999287	1.0	0.111111	0.200000

fit_time          0.324430
score_time        0.029268
test_accuracy     0.999376
test_precision    0.800000
test_recall       0.166667
test_f1           0.266263
dtype: float64

y_pred_proba = baseline_model.predict_proba(X_test)[:, 1]

fpr, tpr, thresholds = roc_curve(y_test, y_pred_proba)
auc = roc_auc_score(y_test, y_pred_proba)

plt.plot(fpr, tpr, label='ROC Curve')
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('ROC Curve')
plt.text(0.805, 0.1, f'AUC = {auc:.2f}')
plt.legend()
plt.show()

The Area Under Receiver Operating Characteristics Curve for the baseline model is 94%.
This means the model is 94% good and robust in classifying the credit card transaction as fraud or non-fraud.

Handling Class Imbalance

Finding class weights

class_weights = compute_class_weight('balanced', classes=y_train.unique(), y=y_train)
class_weights_dict = {0: class_weights[0], 1: class_weights[1]}
class_weights_dict

{0: 0.500369952708455, 1: 676.2620481927711}

Building Balanced Model

balanced_model = LogisticRegression(class_weight=class_weights_dict, random_state=8883828)
balanced_model.fit(X_train, y_train)

LogisticRegression(class_weight={0: 0.500369952708455, 1: 676.2620481927711},
                   random_state=8883828)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

Model Evaluation

cv_results = cross_validate(balanced_model, X_test, y_test, scoring=scoring_metrics, cv=5)
cv_df = pd.DataFrame.from_dict(cv_results)
display(cv_df)
print(cv_df.mean())

	fit_time	score_time	test_accuracy	test_precision	test_recall	test_f1
0	1.007422	0.051835	0.921700	0.006795	0.750000	0.013468
1	0.819167	0.024921	0.921700	0.007910	0.875000	0.015677
2	0.600832	0.030262	0.930340	0.006378	0.625000	0.012626
3	0.789669	0.032063	0.923659	0.009259	0.888889	0.018328
4	0.798304	0.031701	0.947265	0.010084	0.666667	0.019868

fit_time          0.803079
score_time        0.034156
test_accuracy     0.928933
test_precision    0.008085
test_recall       0.761111
test_f1           0.015993
dtype: float64

Unexpectedly, balancing the class is reducing performance for this dataset.
The baseline model has average f1-score 0.266263 whereas the balanced model has only 0.015993.
Also, the precision is very very less 0.008085. So this model is performing very badly for credit card fraud detection.

KNN Model

knn_model = KNeighborsClassifier()
knn_model.fit(X_train, y_train)

KNeighborsClassifier()

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

cv_results = cross_validate(knn_model, X_test, y_test, scoring=scoring_metrics, cv=5)
cv_df = pd.DataFrame.from_dict(cv_results)
display(cv_df)
print(cv_df.mean())

	fit_time	score_time	test_accuracy	test_precision	test_recall	test_f1
0	0.103518	3.074468	0.999466	0.750000	0.375000	0.500
1	0.060120	2.614231	0.999733	0.857143	0.750000	0.800
2	0.053923	2.891817	0.999466	1.000000	0.250000	0.400
3	0.064559	2.722029	0.999466	0.714286	0.555556	0.625
4	0.056105	2.556229	0.999466	1.000000	0.333333	0.500

fit_time          0.067645
score_time        2.771755
test_accuracy     0.999519
test_precision    0.864286
test_recall       0.452778
test_f1           0.565000
dtype: float64

Support Vector Machine

svm_model = SVC(random_state=8883828)
svm_model.fit(X_train, y_train)

SVC(random_state=8883828)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

cv_results = cross_validate(svm_model, X_test, y_test, scoring=scoring_metrics, cv=5)
cv_df = pd.DataFrame.from_dict(cv_results)
display(cv_df)
print(cv_df.mean())

	fit_time	score_time	test_accuracy	test_precision	test_recall	test_f1
0	1.373675	0.801016	0.999466	0.750000	0.375000	0.500000
1	1.384840	0.993112	0.999822	1.000000	0.750000	0.857143
2	1.230355	0.703078	0.999555	1.000000	0.375000	0.545455
3	1.251927	0.716198	0.999466	0.714286	0.555556	0.625000
4	0.986280	0.541178	0.999555	1.000000	0.444444	0.615385

fit_time          1.245415
score_time        0.750917
test_accuracy     0.999572
test_precision    0.892857
test_recall       0.500000
test_f1           0.628596
dtype: float64

Since the dataset is imbalanced, we cannot consider the accuracy as it may skew.
So, considering F1-score, SVM (0.015993) performs better than the baseline Logistic Regression (0.266263) and KNN (0.565000).

For credit card fraud detection model, the precision is more important becuase misclassifying a legitimate transaction as a fraudulent, the customers would be annoyed and it will reduce the reputaion of the bank.
Based on precision also, the SVM (0.892857) outperformed Logistic Regression (0.8) and KNN (0.864286).

---